Screen reader not reading the line of text after using insert text to change color

[SirishaGorasa]

Description of the bug

I have a tagged PDF in which I wanted to change the color of one half of one line of text. Post changing the color using page.insert_text() method, screen reader is not reading the line of text.

Can you help me understand why this had happened and the mechanism to adopt to overcome it?

How to reproduce the bug

import fitz
import json
fitz.TOOLS.set_subset_fontnames(True) 
doc=fitz.open("TagTest_tagged.pdf")
with open('color.json', 'r') as file:
    data = json.load(file)
try:
    page = doc[int(data["page"]-1)]
    bg_color=[cv/255 for cv in data["background_color"]]
    page.draw_rect(fitz.Rect(data["bbox"]), color=bg_color,fill=bg_color) 
    position = data["position"] 
    rect=data["bbox"]
    page.add_redact_annot(rect)  
    page.apply_redactions() 
    fg_color=[cv/255 for cv in data["foreground_color"]]
    page.insert_text(position,text=data["text"],fontsize=data["font_size"],fontname="Times-Roman",color=fg_color)
    doc.save("TagTestNew_tagged.pdf")
except Exception as e:  
    print(e)
finally:
    doc.close()

In the saved document, the content that has been inserted using insert text is not readable by screen reader whereas in the input document, the content is readable.
I am providing the input tagged PDF and the json file.
color.json
TagTest_tagged.pdf

May I know the reason why this had happened? If so, how can I resolve this even after re-insertion of text?

PyMuPDF version

1.23.x or earlier

Operating system

Windows

Python version

3.11

[JorjMcKie]

Cannot reproduce: everything works as designed in current PyMuPDF version. out of command mutool draw -o test.txt TagTestNew_tagged.pdf is this as expected:

Color contrast issue 
 
in tagged PDF

[SirishaGorasa]

Is it working as desired with latest PyMuPDF version? May I know if you've used 1.23.1 to reproduce the same?

One more question is that, Is this the screen reader output text ?

Color contrast issue 
 
in tagged PDF

[JorjMcKie]

As mentioned, the text was produced by MuPDF's mutool (cli), to demonstrate that extractable text has been inserted. The output PDF also looks as expected.

We always only check whether errors occur in the most current version and never fix them in back levels.
Consequently you should always make sure that a supposed bug is reproducible in the current PyMuPDF version.
The current version is 1.25.1.

[SirishaGorasa]

I have tried using PyMuPDF of version 1.25.1 also and noticed that screen reader is not able to read the section of text in green color even using this version. The text looks visually similar but not while using NVDA screen reader to read.

[JorjMcKie]

Well, all the PDF viewers I am using, including Adobe Acrobat show the PDF as expected and also produce the expected result upon copy-paste. The usual text extractors like XPDF/pdftotext also produce the right output.
So this is sufficient assertion for correct functioning.
You probably should file a bug report at NVDA screen reader - whoever they are.

[SirishaGorasa]

Let me make one thing a bit more clear. Screen readers like NVDA read the content not based on visual representation on the PDF page but instead use associated tag structure. In this case, I noticed that the tag tree looks like below before inserting text in green color.

After inserting text with green color as the foreground color, the tag tree is looking like below

You can make out a clear difference that the text in tagged PDF which originally is under <Span> tag in the original PDF's tag tree has now disappeared, which is not allowing screen reader like NVDA to read the content.

[JorjMcKie]

Ah, understand you now, thanks.
Well, all PyMuPDF insertions (text, drawings, images, ...) do not support / maintain tags.
So the behavior you saw is inevitable and there is no way to do anything about it.

[SirishaGorasa]

Oh, Is it? Even I got the point now. Thanks for the clarification.

Do you have an idea of any alternate workarounds to ensure tag even after changing color of the existing text?

[JorjMcKie]

Sorry, I should have added an announcement for moving this thing to Discussions.
Anyway.
What you can try is:

1. Do the insertion(s) you need.
2. Make a new empty PDF
3. Insert the original PDF into the new one using .insert_pdf().
   The resulting PDF will have all tagging information removed and be a "normal" PDF without tags.

Then try and see if you obscure NVDA screen reader can deal with that resulting PDF.